CAREER: DURAIS : A Platform for Co-Designing and Understanding the Roles of Conversational Artificial Intelligence Systems on Caregiving

该奖项的部分资金来自《2021年美国救援计划法案》(公法117-2)。照顾包括社交、情感和身体上要求很高的工作。然而，照顾者的资源是有限的。本研究项目旨在推进以人为中心的人工智能(AI)的最新发展。这个项目调查了对话式人工智能系统(CAI)在支持照顾者(例如，父母、监护人、其他家庭成员)方面可以发挥的作用，以及领域专家和照顾者如何共同努力来定义这些角色。通过研究照看的两种不同环境--养育和痴呆症生活--这项研究有助于：(1)了解对话型人工智能在解决照看需求方面的作用和限制；以及(2)为利益相关者开发一个平台，以共同设计支持照看工作的对话式人工智能。对话式人工智能有望帮助护理者专注于无法自动化的工作，减少护理需求并提高质量。在整个研究计划中与社区接触有望产生有益的结果。参与的过程包括面向社区的研讨会和研讨会，传播知识，创造与护理、营养和家庭健康等跨学科领域的协同合作机会，通过人工智能技术解决社会的重大问题。该研究项目通过与领域专家和照顾者共同设计来定义这些角色，从而有助于确定人工智能系统如何支持与照看有关的工作的社会和技术边界(例如，养育父母、照顾生病的人)。该项目首先对照顾者支持活动的对话数据源进行形成性访谈和分析，以进行需求工程，在这些需求工程上，照顾者工作可以通过计算机辅助教学得到支持、减少或自动化。用生成式和基于规则的计算机辅助教学来实现这些要求，然后测试各种语言模型在有效地生成自动计算机辅助教学反应方面的比较性能。在这项形成性工作的基础上，该项目开发了Durais平台，允许领域专家和照顾者共同设计支持照看的计算机辅助教学。通过开发和评估两种不同环境下的计算机辅助教学协同设计平台--育儿和痴呆症生活--这项研究有助于测试计算机辅助教学协同设计如何解决基于规则的系统效率低下和生成模型不可靠的挑战，并评估自动生成的生成性计算机辅助教学对话与基于规则的计算机辅助教学如何能够独特地支持护理。该研究项目有助于理解共同设计人工智能系统的新范式，以支持社会和文化上的复杂和苛刻的工作。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

This award is funded in part under the American Rescue Plan Act of 2021 (Public Law 117-2). Caregiving involves socially, emotionally, and physically demanding work. However, resources for caregivers are limited. This research project aims to advance the state-of-the-art of human-centered artificial intelligence (AI). This project investigate the roles conversational AI systems (CAI) can play in supporting caregivers (e.g., parents, guardians, other family members) and how domain experts and caregivers can work together to define these roles. By studying two different contexts of caregiving--parenting and living with dementia--the research contributes to; (1) understanding the roles and limitations on when conversational AI addresses caregiving needs; and, (2) developing a platform for stakeholders to co-design conversational AI that supports the work of caregiving. Conversational AI is expected to help caregivers to focus on the work that cannot be automated, reducing caregiving demands and improving quality. Engaging with community throughout the research program is expected to generate beneficial outcomes. Processes of engagement include community-facing workshops and seminars, disseminating knowledge, and generating synergistic collaborative opportunities with cross-disciplinary fields, such as nursing, nutrition, and family health, solving society’s big problems through AI technologies. The research project contributes to identifying the social and technical boundaries of how AI systems can support the work involved in caregiving (e.g., parenting, caregiving ill individuals) by defining these roles through co-design with domain experts and caregivers. The project begins by performing formative interviews and analysis of conversational data sources of caregiver support activities to conduct requirements engineering on which caregiving work can be supported, reduced, or automated through CAI. Implementing these requirements with generative and rule-based CAI then tests comparative performances of various language models in effectively generating automated CAI responses. Building on this formative work, the project develops the DURAIS platform, allowing domain experts and caregivers to co-design CAI supporting caregiving. By developing and evaluating a co-design platform of CAI in two different contexts of caregiving—parenting and living with dementia--the research contributes to testing how CAI co-design can resolve challenges around inefficiencies of rule-based systems and unreliability of generative models and evaluating how automatically generated conversations of generative- vs. rule-based CAI can uniquely support caregiving. The research program broadly contributes to understanding novel paradigms for co-designing AI systems that support socially and culturally complex and demanding work.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.